xref: /csrg-svn/sys/net/route.c (revision 68025)
1 /*
2  * Copyright (c) 1980, 1986, 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * %sccs.include.redist.c%
6  *
7  *	@(#)route.c	8.2.1.1 (Berkeley) 12/02/94
8  */
9 
10 #include <sys/param.h>
11 #include <sys/systm.h>
12 #include <sys/proc.h>
13 #include <sys/mbuf.h>
14 #include <sys/socket.h>
15 #include <sys/socketvar.h>
16 #include <sys/domain.h>
17 #include <sys/protosw.h>
18 #include <sys/ioctl.h>
19 
20 #include <net/if.h>
21 #include <net/route.h>
22 #include <net/raw_cb.h>
23 
24 #include <netinet/in.h>
25 #include <netinet/in_var.h>
26 
27 #ifdef NS
28 #include <netns/ns.h>
29 #endif
30 
31 #define	SA(p) ((struct sockaddr *)(p))
32 
33 int	rttrash;		/* routes not in table but not freed */
34 struct	sockaddr wildcard;	/* zero valued cookie for wildcard searches */
35 
36 void
37 rtable_init(table)
38 	void **table;
39 {
40 	struct domain *dom;
41 	for (dom = domains; dom; dom = dom->dom_next)
42 		if (dom->dom_rtattach)
43 			dom->dom_rtattach(&table[dom->dom_family],
44 			    dom->dom_rtoffset);
45 }
46 
47 void
48 route_init()
49 {
50 	rn_init();	/* initialize all zeroes, all ones, mask table */
51 	rtable_init((void **)rt_tables);
52 }
53 
54 /*
55  * Packet routing routines.
56  */
57 void
58 rtalloc(ro)
59 	register struct route *ro;
60 {
61 	if (ro->ro_rt && ro->ro_rt->rt_ifp && (ro->ro_rt->rt_flags & RTF_UP))
62 		return;				 /* XXX */
63 	ro->ro_rt = rtalloc1(&ro->ro_dst, 1);
64 }
65 
66 struct rtentry *
67 rtalloc1(dst, report)
68 	register struct sockaddr *dst;
69 	int report;
70 {
71 	register struct radix_node_head *rnh = rt_tables[dst->sa_family];
72 	register struct rtentry *rt;
73 	register struct radix_node *rn;
74 	struct rtentry *newrt = 0;
75 	struct rt_addrinfo info;
76 	int  s = splnet(), err = 0, msgtype = RTM_MISS;
77 
78 	bzero((caddr_t)&info, sizeof(info));
79 	info.rti_info[RTAX_DST] = dst;
80 	if (rnh && (rn = rnh->rnh_matchaddr((caddr_t)dst, rnh)) &&
81 	    ((rn->rn_flags & RNF_ROOT) == 0)) {
82 		newrt = rt = (struct rtentry *)rn;
83 		if (report && (rt->rt_flags & RTF_CLONING)) {
84 			info.rti_ifa = rt->rt_ifa;
85 			info.rti_ifp = rt->rt_ifp;
86 			info.rti_flags = rt->rt_flags & ~RTF_CLONING;
87 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
88 			if ((info.rti_info[RTAX_NETMASK] = rt->rt_genmask) == 0)
89 				info.rti_flags |= RTF_HOST;
90 			err = rtrequest1(RTM_RESOLVE, &info, &newrt);
91 			if (err) {
92 				newrt = rt;
93 				rt->rt_refcnt++;
94 				goto miss;
95 			}
96 			newrt->rt_rmx = rt->rt_rmx;
97 			if (rt->rt_flags & RTF_XRESOLVE) {
98 				msgtype = RTM_RESOLVE;
99 				goto miss;
100 			}
101 		} else
102 			rt->rt_refcnt++;
103 	} else {
104 		rtstat.rts_unreach++;
105 	miss:	if (report) {
106 			rt_missmsg(msgtype, &info, 0, err);
107 		}
108 	}
109 	splx(s);
110 	return (newrt);
111 }
112 
113 void
114 rtfree(rt)
115 	register struct rtentry *rt;
116 {
117 	register struct ifaddr *ifa;
118 
119 	if (rt == 0)
120 		panic("rtfree");
121 	rt->rt_refcnt--;
122 	if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_UP) == 0) {
123 		if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
124 			panic ("rtfree 2");
125 		rttrash--;
126 		if (rt->rt_refcnt < 0) {
127 			printf("rtfree: %x not freed (neg refs)\n", rt);
128 			return;
129 		}
130 		ifa = rt->rt_ifa;
131 		IFAFREE(ifa);
132 		Free(rt_key(rt));
133 		Free(rt);
134 	}
135 }
136 
137 void
138 ifafree(ifa)
139 	register struct ifaddr *ifa;
140 {
141 	if (ifa == NULL)
142 		panic("ifafree");
143 	if (ifa->ifa_refcnt == 0)
144 		free(ifa, M_IFADDR);
145 	else
146 		ifa->ifa_refcnt--;
147 }
148 
149 /*
150  * Force a routing table entry to the specified
151  * destination to go through the given gateway.
152  * Normally called as a result of a routing redirect
153  * message from the network layer.
154  *
155  * N.B.: must be called at splnet
156  *
157  */
158 int
159 rtredirect(dst, gateway, netmask, flags, src, rtp)
160 	struct sockaddr *dst, *gateway, *netmask, *src;
161 	int flags;
162 	struct rtentry **rtp;
163 {
164 	register struct rtentry *rt;
165 	int error = 0;
166 	short *stat = 0;
167 	struct rt_addrinfo info;
168 	struct ifaddr *ifa;
169 
170 	/* verify the gateway is directly reachable */
171 	if ((ifa = ifa_ifwithnet(gateway)) == 0) {
172 		error = ENETUNREACH;
173 		goto out;
174 	}
175 	rt = rtalloc1(dst, 0);
176 	/*
177 	 * If the redirect isn't from our current router for this dst,
178 	 * it's either old or wrong.  If it redirects us to ourselves,
179 	 * we have a routing loop, perhaps as a result of an interface
180 	 * going down recently.
181 	 */
182 #define	equal(a1, a2) (bcmp((caddr_t)(a1), (caddr_t)(a2), (a1)->sa_len) == 0)
183 	if (!(flags & RTF_DONE) && rt &&
184 	     (!equal(src, rt->rt_gateway) || rt->rt_ifa != ifa))
185 		error = EINVAL;
186 	else if (ifa_ifwithaddr(gateway))
187 		error = EHOSTUNREACH;
188 	if (error)
189 		goto done;
190 	/*
191 	 * Create a new entry if we just got back a wildcard entry
192 	 * or the the lookup failed.  This is necessary for hosts
193 	 * which use routing redirects generated by smart gateways
194 	 * to dynamically build the routing tables.
195 	 */
196 	if ((rt == 0) || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
197 		goto create;
198 	/*
199 	 * Don't listen to the redirect if it's
200 	 * for a route to an interface.
201 	 */
202 	if (rt->rt_flags & RTF_GATEWAY) {
203 		if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
204 			/*
205 			 * Changing from route to net => route to host.
206 			 * Create new route, rather than smashing route to net.
207 			 */
208 		create:
209 			flags |=  RTF_GATEWAY | RTF_DYNAMIC;
210 			error = rtrequest((int)RTM_ADD, dst, gateway,
211 				    netmask, flags,
212 				    (struct rtentry **)0);
213 			stat = &rtstat.rts_dynamic;
214 		} else {
215 			/*
216 			 * Smash the current notion of the gateway to
217 			 * this destination.  Should check about netmask!!!
218 			 */
219 			rt->rt_flags |= RTF_MODIFIED;
220 			flags |= RTF_MODIFIED;
221 			stat = &rtstat.rts_newgateway;
222 			rt_setgate(rt, rt_key(rt), gateway);
223 		}
224 	} else
225 		error = EHOSTUNREACH;
226 done:
227 	if (rt) {
228 		if (rtp && !error)
229 			*rtp = rt;
230 		else
231 			rtfree(rt);
232 	}
233 out:
234 	if (error)
235 		rtstat.rts_badredirect++;
236 	else if (stat != NULL)
237 		(*stat)++;
238 	bzero((caddr_t)&info, sizeof(info));
239 	info.rti_info[RTAX_DST] = dst;
240 	info.rti_info[RTAX_GATEWAY] = gateway;
241 	info.rti_info[RTAX_NETMASK] = netmask;
242 	info.rti_info[RTAX_AUTHOR] = src;
243 	rt_missmsg(RTM_REDIRECT, &info, flags, error);
244 }
245 
246 /*
247 * Routing table ioctl interface.
248 */
249 int
250 rtioctl(req, data, p)
251 	int req;
252 	caddr_t data;
253 	struct proc *p;
254 {
255 	return (EOPNOTSUPP);
256 }
257 
258 struct ifaddr *
259 ifa_ifwithroute(flags, dst, gateway)
260 	int flags;
261 	struct sockaddr	*dst, *gateway;
262 {
263 	register struct ifaddr *ifa;
264 	if ((flags & RTF_GATEWAY) == 0) {
265 		/*
266 		 * If we are adding a route to an interface,
267 		 * and the interface is a pt to pt link
268 		 * we should search for the destination
269 		 * as our clue to the interface.  Otherwise
270 		 * we can use the local address.
271 		 */
272 		ifa = 0;
273 		if (flags & RTF_HOST)
274 			ifa = ifa_ifwithdstaddr(dst);
275 		if (ifa == 0)
276 			ifa = ifa_ifwithaddr(gateway);
277 	} else {
278 		/*
279 		 * If we are adding a route to a remote net
280 		 * or host, the gateway may still be on the
281 		 * other end of a pt to pt link.
282 		 */
283 		ifa = ifa_ifwithdstaddr(gateway);
284 	}
285 	if (ifa == 0)
286 		ifa = ifa_ifwithnet(gateway);
287 	if (ifa == 0) {
288 		struct rtentry *rt = rtalloc1(dst, 0);
289 		if (rt == 0)
290 			return (0);
291 		rt->rt_refcnt--;
292 		if ((ifa = rt->rt_ifa) == 0)
293 			return (0);
294 	}
295 	if (ifa->ifa_addr->sa_family != dst->sa_family) {
296 		struct ifaddr *oifa = ifa;
297 		ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
298 		if (ifa == 0)
299 			ifa = oifa;
300 	}
301 	return (ifa);
302 }
303 
304 int
305 rtgateinfo(info)
306 	register struct rt_addrinfo *info;
307 {
308 	struct ifaddr *ifa;
309 	struct ifnet *ifp;
310 	int error = 0;
311 /* Sleazy use of local variables throughout this function and next, XXX!!!! */
312 #define dst	info->rti_info[RTAX_DST]
313 #define gateway	info->rti_info[RTAX_GATEWAY]
314 #define netmask	info->rti_info[RTAX_NETMASK]
315 #define ifaaddr	info->rti_info[RTAX_IFA]
316 #define ifpaddr	info->rti_info[RTAX_IFP]
317 #define flags	info->rti_flags
318 
319 	/* ifp may be specified by sockaddr_dl
320 		   when protcol address is ambiguous */
321 	if (info->rti_ifp == 0 && ifpaddr) {
322 		ifa = ifa_ifwithnet(ifpaddr);
323 		info->rti_ifp = ifa ? ifa->ifa_ifp : 0;
324 	}
325 	if (info->rti_ifa == 0) {
326 		struct sockaddr *sa;
327 		if (ifp = info->rti_ifp) {
328 			sa = ifaaddr ? ifaaddr : (gateway ? gateway : dst);
329 			info->rti_ifa = ifaof_ifpforaddr(sa, ifp);
330 		} else
331 			info->rti_ifa = ifa_ifwithroute(flags, dst, gateway);
332 	}
333 	if (ifa = info->rti_ifa) {
334 		if (info->rti_ifp == 0)
335 			info->rti_ifp = ifa->ifa_ifp;
336 		if (gateway == 0)
337 		    gateway = ifa->ifa_addr;
338 	} else
339 		error = ENETUNREACH;
340 	return error;
341 }
342 
343 
344 #define ROUNDUP(a) (a>0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
345 
346 int
347 rtrequest1(req, info, ret_nrt)
348 	int req;
349 	register struct rt_addrinfo *info;
350 	struct rtentry **ret_nrt;
351 {
352 	int s = splnet(); int error = 0;
353 	register struct rtentry *rt;
354 	register struct radix_node *rn;
355 	register struct radix_node_head *rnh;
356 	struct sockaddr *ndst;
357 	struct ifaddr *ifa = info->rti_ifa;
358 	struct ifnet *ifp;
359 #define senderr(x) { error = x ; goto bad; }
360 #define RTE(x) ((struct rtentry *)x)
361 
362 	if ((rnh = rt_tables[dst->sa_family]) == 0)
363 		senderr(ESRCH);
364 	if (flags & RTF_HOST)
365 		netmask = 0;
366 	switch (req) {
367 	case RTM_DELETE:
368 		if ((ifaaddr || ifpaddr) && rnh->rnh_delete1) {
369 			extern struct radix_node_head *mask_rnhead;
370 			if (error = rtgateinfo(info))
371 				senderr(error);
372 			ifa = info->rti_ifa;
373 			ifp = info->rti_ifp;
374 			if (netmask && (rn = rn_search(netmask,
375                                             mask_rnhead->rnh_treetop)))
376                                 netmask = SA(rn->rn_key);
377 			for (rn = rnh->rnh_matchaddr(dst, rnh); rn;
378 						    rn = rn->rn_dupedkey) {
379 				/* following test includes case netmask == 0 */
380 				if (SA(rn->rn_mask) == netmask &&
381 				    ((ifaaddr && ifa == RTE(rn)->rt_ifa) ||
382 				     (!ifaaddr && ifp == RTE(rn)->rt_ifp)))
383 					break;
384 			if (rn == 0 || (rn = rnh->rnh_delete1(rn, rnh)) == 0)
385 				senderr(ESRCH);
386 			}
387 		} else if ((rn = rnh->rnh_deladdr(dst, netmask, rnh)) == 0)
388 			senderr(ESRCH);
389 		if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
390 			panic ("rtrequest delete");
391 		rt = RTE(rn);
392 		rt->rt_flags &= ~RTF_UP;
393 		if (rt->rt_gwroute) {
394 			rt = rt->rt_gwroute; RTFREE(rt);
395 			(rt = RTE(rn))->rt_gwroute = 0;
396 		}
397 		if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest)
398 			ifa->ifa_rtrequest(RTM_DELETE, rt, info);
399 		rttrash++;
400 		if (ret_nrt)
401 			*ret_nrt = rt;
402 		else if (rt->rt_refcnt <= 0) {
403 			rt->rt_refcnt++;
404 			rtfree(rt);
405 		}
406 		break;
407 
408 
409 	case RTM_ADD:
410 		if (ifa == 0) {
411 			if (error = rtgateinfo(info))
412 				senderr(error);
413 			ifa = info->rti_ifa;
414 		}
415 	case RTM_RESOLVE:
416 		R_Malloc(rt, struct rtentry *, sizeof(*rt));
417 		if (rt == 0)
418 			senderr(ENOBUFS);
419 		Bzero(rt, sizeof(*rt));
420 		rt->rt_flags = RTF_UP | flags;
421 		if (rt_setgate(rt, dst, gateway)) {
422 			Free(rt);
423 			senderr(ENOBUFS);
424 		}
425 		ndst = rt_key(rt);
426 		if (netmask) {
427 			rt_maskedcopy(dst, ndst, netmask);
428 		} else
429 			Bcopy(dst, ndst, dst->sa_len);
430 		rn = rnh->rnh_addaddr((caddr_t)ndst, (caddr_t)netmask,
431 					rnh, rt->rt_nodes);
432 		if (rn == 0) {
433 			if (rt->rt_gwroute)
434 				rtfree(rt->rt_gwroute);
435 			Free(rt_key(rt));
436 			Free(rt);
437 			senderr(EEXIST);
438 		}
439 		ifa->ifa_refcnt++;
440 		rt->rt_ifa = ifa;
441 		rt->rt_ifp = ifa->ifa_ifp;
442 		if (ifa->ifa_rtrequest)
443 			ifa->ifa_rtrequest(req, rt, info);
444 		if (ret_nrt) {
445 			*ret_nrt = rt;
446 			rt->rt_refcnt++;
447 		}
448 		break;
449 	}
450 bad:
451 	splx(s);
452 	return (error);
453 #undef dst
454 #undef gateway
455 #undef netmask
456 #undef ifaaddr
457 #undef ifpaddr
458 #undef flags
459 }
460 
461 int
462 rtrequest(req, dst, gateway, netmask, flags, ret_nrt)
463 	int req, flags;
464 	struct sockaddr *dst, *gateway, *netmask;
465 	struct rtentry **ret_nrt;
466 {
467 	struct rt_addrinfo info;
468 
469 	bzero((caddr_t)&info, sizeof(info));
470 	info.rti_flags = flags;
471 	info.rti_info[RTAX_DST] = dst;
472 	info.rti_info[RTAX_GATEWAY] = gateway;
473 	info.rti_info[RTAX_NETMASK] = netmask;
474 	return rtrequest1(req, info, ret_nrt);
475 }
476 
477 int
478 rt_setgate(rt0, dst, gate)
479 	struct rtentry *rt0;
480 	struct sockaddr *dst, *gate;
481 {
482 	caddr_t new, old;
483 	int dlen = ROUNDUP(dst->sa_len), glen = ROUNDUP(gate->sa_len);
484 	register struct rtentry *rt = rt0;
485 
486 	if (rt->rt_gateway == 0 || glen > ROUNDUP(rt->rt_gateway->sa_len)) {
487 		old = (caddr_t)rt_key(rt);
488 		R_Malloc(new, caddr_t, dlen + glen);
489 		if (new == 0)
490 			return 1;
491 		rt->rt_nodes->rn_key = new;
492 	} else {
493 		new = rt->rt_nodes->rn_key;
494 		old = 0;
495 	}
496 	Bcopy(gate, (rt->rt_gateway = (struct sockaddr *)(new + dlen)), glen);
497 	if (old) {
498 		Bcopy(dst, new, dlen);
499 		Free(old);
500 	}
501 	if (rt->rt_gwroute) {
502 		rt = rt->rt_gwroute; RTFREE(rt);
503 		rt = rt0; rt->rt_gwroute = 0;
504 	}
505 	if (rt->rt_flags & RTF_GATEWAY) {
506 		rt->rt_gwroute = rtalloc1(gate, 1);
507 	}
508 	return 0;
509 }
510 
511 void
512 rt_maskedcopy(src, dst, netmask)
513 	struct sockaddr *src, *dst, *netmask;
514 {
515 	register u_char *cp1 = (u_char *)src;
516 	register u_char *cp2 = (u_char *)dst;
517 	register u_char *cp3 = (u_char *)netmask;
518 	u_char *cplim = cp2 + *cp3;
519 	u_char *cplim2 = cp2 + *cp1;
520 
521 	*cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
522 	cp3 += 2;
523 	if (cplim > cplim2)
524 		cplim = cplim2;
525 	while (cp2 < cplim)
526 		*cp2++ = *cp1++ & *cp3++;
527 	if (cp2 < cplim2)
528 		bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2));
529 }
530 
531 /*
532  * Set up a routing table entry, normally
533  * for an interface.
534  */
535 int
536 rtinit(ifa, cmd, flags)
537 	register struct ifaddr *ifa;
538 	int cmd, flags;
539 {
540 	register struct rtentry *rt;
541 	register struct sockaddr *dst;
542 	register struct sockaddr *deldst;
543 	struct mbuf *m = 0;
544 	struct rtentry *nrt = 0;
545 	int error;
546 
547 	dst = flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
548 	if (cmd == RTM_DELETE) {
549 		if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
550 			m = m_get(M_WAIT, MT_SONAME);
551 			deldst = mtod(m, struct sockaddr *);
552 			rt_maskedcopy(dst, deldst, ifa->ifa_netmask);
553 			dst = deldst;
554 		}
555 		if (rt = rtalloc1(dst, 0)) {
556 			rt->rt_refcnt--;
557 			if (rt->rt_ifa != ifa) {
558 				if (m)
559 					(void) m_free(m);
560 				return (flags & RTF_HOST ? EHOSTUNREACH
561 							: ENETUNREACH);
562 			}
563 		}
564 	}
565 	error = rtrequest(cmd, dst, ifa->ifa_addr, ifa->ifa_netmask,
566 			flags | ifa->ifa_flags, &nrt);
567 	if (m)
568 		(void) m_free(m);
569 	if (cmd == RTM_DELETE && error == 0 && (rt = nrt)) {
570 		rt_newaddrmsg(cmd, ifa, error, nrt);
571 		if (rt->rt_refcnt <= 0) {
572 			rt->rt_refcnt++;
573 			rtfree(rt);
574 		}
575 	}
576 	if (cmd == RTM_ADD && error == 0 && (rt = nrt)) {
577 		rt->rt_refcnt--;
578 		if (rt->rt_ifa != ifa) {
579 			printf("rtinit: wrong ifa (%x) was (%x)\n", ifa,
580 				rt->rt_ifa);
581 			if (rt->rt_ifa->ifa_rtrequest)
582 			    rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt, SA(0));
583 			IFAFREE(rt->rt_ifa);
584 			rt->rt_ifa = ifa;
585 			rt->rt_ifp = ifa->ifa_ifp;
586 			ifa->ifa_refcnt++;
587 			if (ifa->ifa_rtrequest)
588 			    ifa->ifa_rtrequest(RTM_ADD, rt, SA(0));
589 		}
590 		rt_newaddrmsg(cmd, ifa, error, nrt);
591 	}
592 	return (error);
593 }
594